We present evidence from a 1-year longitudinal study, using information from the Stroke Recovery in Underserved Population (SRUP) database, for the psychometrics of Center for Epidemiologic Studies Depression Scale (CES-D) in patients after stroke. This study included a sample of 828 adults, 55 years old and older with a first-time stroke, and admitted to one of eleven inpatient rehabilitation facilities in the United States. Confirmatory factor analyses indicated four-factor structure of the CES-D. Classical test theory and Rasch analyses further supported the reliability and validity of the CES-D. This measure allows investigators and clinicians to better characterize depressive symptoms in patients after stroke.
To examine the psychometrics of the Center for Epidemiologic Studies Depression (CES-D) for patients after stroke using both classical test theory (CTT) and Rasch analyses.
A secondary data analysis of the Stroke Recovery in Underserved Population database (2005-2006), which was a prospective observational study with post-stroke patients.
Eleven inpatient rehabilitation facilities located across diverse regions of the United States.
Participants (or Animals, Specimens, Cadavers):
Patients (N=828) who received inpatient stroke rehabilitation and had complete follow-up data were included in the analysis.
Main Outcome Measure(s):
CES-D was administrated at discharge, 3-month, and 12-month follow-up.
Comparisons of competing models using confirmatory factor analyses indicated that the four-factor model (somatic symptoms, depressed affect, positive affect, and interpersonal difficulties) possessed the best model fit (CFI=0.98; TLI= 0.98; RMSEA=0.05). CTT analyses showed that the CES-D had satisfactory internal consistency (α=0.74-0.87). Rasch analyses supported the unidimensionality of each CES-D subscale with no misfit item. Rating scale analyses indicated the need to collapse response categories 2 and 3 of the somatic symptoms subscale. Reliability analyses indicated satisfactory item reliability but poor person separation of all subscales. A floor effect for persons and an uneven spread of items across the spectrum were observed. A total of 3 differential item functioning (DIF) items (depressed affect: 1; somatic symptoms: 2) across gender, and 1 DIF item (depressed affect: 1) across the time course of recovery were found but none of them reached practical relevance. All items of the somatic symptoms subscale indicated DIF across time.
We confirmed the four-factor structure of the CES-D, supporting the use of four subscales to characterize depressive symptoms in patients after stroke. We verified the psychometrics of the CES-D and reported its DIF items. Other supplementary instruments are needed when comparing somatic symptoms across the time course of recovery.